Decentralized ComBat-Mega for Harmonizing FNC Data in Coinstac
Data harmonization in neuroimaging helps in standardizing the data collected from different research facilities to remove site effects and improve statistical analysis performed on such data, which may incur additional source dependency. In this work, we propose a decentralized comBat-Mega algorithm which interpolates missing data using linear regression of co-variates at the local sites and performs data harmonization on functional network connectivity (FNC) matrices using decen-tralization approach. We verify our model by using data collected by two different cohorts where our model performs on par with the centralized model while using only information from locally trained models without any data sharing.
Authors
- Sunitha Basodi (ORCID: https://orcid.org/0000-0003-3453-3668)
- Biozid Bostami (ORCID: https://orcid.org/0000-0001-7771-2581)
- Sandeep Panta (ORCID: https://orcid.org/0000-0002-6183-9139)
- Vince D. Calhoun (ORCID: https://orcid.org/0000-0001-9058-0747)
- Frank Gerard Hillary (ORCID: https://orcid.org/0000-0002-1427-0218)
- Charles A. Ellis (ORCID: https://orcid.org/0000-0002-1547-9996)
Institutions
- Georgia Institute of Technology (US)
- Pennsylvania State University (US)
- University Medical Center Groningen (NL)
- Emory University (US)
- University of Groningen (NL)
- Georgia State University (US)
- Center for Translational Research in Neuroimaging and Data Science (US)
Publication Details
- Published
- 2025-04-14
- DOI
- https://doi.org/10.1109/isbi60581.2025.10980680
- Primary Topic
- Seismic Imaging and Inversion Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00